Decentralized Waveform Co-Design for Integrated Sensing and Communications Systems via Approximate Dynamic Programming

The increasing demand for cost-efficient yet reliable swarms of multi-function unmanned aerial systems (UASs) has tremendous potential in both military and civilian applications. Advancements in ISACs research, such as RF Convergence exclusively provide a promising path toward implementing multi-function UASs, often limited by insufficient front-end capabilities as the network scales up. Because the interconnection bandwidth and computational complexity required to process the data among the ISACs nodes using centralized architectures are high. In this paper, we propose a decentralized waveform co-design method to reduce computational complexity while maximizing the mutual benefits of users in a multi-function UASs network based on the theory of decentralized, partially-observable Markov decision processes (Dec-POMDPs). To address the computational intractability of solving Dec-POMDPs (as with any decision-theoretic framework), we extend an approximate dynamic programming approach we recently developed-nominal belief-state optimization (NBO) in the context of radar-communications waveform co-design. We conduct a numerical study to benchmark the performance of the Dec-POMDP-based waveform design approach against a centralized decision optimization approach, which demonstrates a 50% reduction in computation time at the cost of moderate loss in ranging precision.